SPSSX Discussion

split plot

Classic

List

Threaded

3 messages Options

julia-12

split plot

Dear all,

I need help with a pair of split plot analysis.

In the first case, I have a three factors complete randomized split-plot
design, with one factor on whole-plot and two factors on sub-plots.

In the second case, is also a three factors complete randomized split-
plot, with two factors on whole-plot and one factor on sub-plot.

I would like to know if SPSS can perform this kind of analysis via Syntax
commands and how it could be done. I would be really very grateful of any
kind of help or comment.

Julia.
PhD Student at University of Santiago de Compostela. Spain.

Marta García-Granero

Re: split plot

Hola Julia:

Saludos desde Pamplona...

I have a worked example of a split-plot design but, unfortunately, a
bit simpler than your layouts (it has only two factors, one in main
plot an the other in sub-plots). You can use it as a starting point to
develop your syntax. With UNIANOVA, they key is using the /TEST
subcommand. It always works, while the approach using MIXED did not in
some particular examples I tried.

If you need more information or help, please send a small sample of
your data and I will try to adjust the syntax for your particular
situation.

*************************************************************
* SPLIT-PLOT ANALYSIS USING UNIANOVA *
* AND MIXED (using EMEANS+SIDAK formultiple comparisons) *
* Split-plot example taken from Irristat 4.4 manual: *
* http://www.irri.org/science/software/irristat.asp *
*************************************************************
* RESPONSE OF MAIZE TO NITROGEN FERTILIZER *
* WHEN IT IS GROWN AFTER DIFFERENT CROPS *
* TALLEYRAND, H. NCRE CEREALS AGRONOMIST, IRA/GAROUA *
* THE TRIAL WAS A SPLIT PLOT DESIGN WITH PRECEEDING CROP AS *
* MAIN PLOT FACTOR AND FOUR LEVELS OF NITROGEN AS SUB-PLOT *
* FACTOR: 0, 45, 90, 135 KG N/ha. MAIZE VARIETY CMS8501. *
*************************************************************

* Dataset *.
DATA LIST FREE/block pcrop nitrogen (3 F4.0) yield (F4.2).
BEGIN DATA
1 5 1 2.35 1 5 2 3.29 1 5 3 4.75 1 5 4 6.39 1 1 1 3.76 1 1 2 5.17
1 1 3 6.30 1 1 4 6.82 1 3 1 2.26 1 3 2 5.17 1 3 3 5.88 1 3 4 6.82
1 2 1 3.71 1 2 2 4.70 1 2 3 5.88 1 2 4 6.20 1 4 1 1.41 1 4 2 3.48
1 4 3 4.61 1 4 4 4.70 2 5 1 4.20 2 5 2 5.23 2 5 3 6.64 2 5 4 6.73
2 1 1 5.28 2 1 2 5.84 2 1 3 6.36 2 1 4 7.39 2 3 1 3.49 2 3 2 5.84
2 3 3 6.88 2 3 4 6.78 2 2 1 4.85 2 2 2 5.98 2 2 3 6.73 2 2 4 7.49
2 4 1 3.12 2 4 2 3.59 2 4 3 4.67 2 4 4 6.50 3 5 1 5.34 3 5 2 6.79
3 5 3 7.31 3 5 4 8.20 3 1 1 6.75 3 1 2 8.11 3 1 3 8.30 3 1 4 8.91
3 3 1 4.26 3 3 2 7.17 3 3 3 7.59 3 3 4 8.34 3 2 1 5.57 3 2 2 5.81
3 2 3 6.23 3 2 4 7.69 3 4 1 4.49 3 4 2 5.67 3 4 3 6.93 3 4 4 7.17
4 5 1 3.38 4 5 2 3.85 4 5 3 5.17 4 5 4 5.36 4 1 1 4.18 4 1 2 5.88
4 1 3 6.30 4 1 4 7.14 4 3 1 2.82 4 3 2 4.61 4 3 3 5.73 4 3 4 5.88
4 2 1 4.47 4 2 2 5.17 4 2 3 5.73 4 2 4 6.34 4 4 1 2.73 4 4 2 3.06
4 4 3 4.70 4 4 4 5.36
END DATA.

VALUE LABEL pcrop 1'PPEAS' 2'CROTL' 3'GNUTS' 4'CWPEA' 5'MAIZE'.
VALUE LABEL nitrogen 1'0' 2'45' 3'90' 4'135'.

* Using UNIANOVA *

UNIANOVA
yield BY block pcrop nitrogen
/RANDOM = block
/METHOD = SSTYPE(3)
/INTERCEPT = EXCLUDE
/TEST= pcrop vs block*pcrop
/DESIGN = pcrop block block*pcrop nitrogen pcrop*nitrogen .

* With MIXED, with a post-hoc analysis for main effects *.

MIXED
yield BY pcrop block nitrogen
/FIXED = pcrop nitrogen nitrogen*pcrop | NOINT
/METHOD = REML
/RANDOM = block block*pcrop | COVTYPE(VC)
/SAVE = RESID
/EMMEANS = TABLES(pcrop) COMPARE ADJ(SIDAK)
/EMMEANS = TABLES(nitrogen) COMPARE ADJ(SIDAK)

j> I need help with a pair of split plot analysis.

j> In the first case, I have a three factors complete randomized split-plot
j> design, with one factor on whole-plot and two factors on sub-plots.

j> In the second case, is also a three factors complete randomized split-
j> plot, with two factors on whole-plot and one factor on sub-plot.

j> I would like to know if SPSS can perform this kind of analysis via Syntax
j> commands and how it could be done. I would be really very grateful of any
j> kind of help or comment.

j> Julia.
j> PhD Student at University of Santiago de Compostela. Spain.

--
Regards,
Dr. Marta García-Granero,PhD mailto:[hidden email]
Statistician
University of Navarra
SPAIN

---
"It is unwise to use a statistical procedure whose use one does
not understand. SPSS syntax guide cannot supply this knowledge, and it
is certainly no substitute for the basic understanding of statistics
and statistical thinking that is essential for the wise choice of
methods and the correct interpretation of their results".

(Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)

LUCINDA M TEAR

/EMMEANS

I am not certain how to explain the type of post hoc tests that are
conducted with the /EMMEANS subcommand. Are they essentially multiple,
one-way ANOVAs using the factors listed on the /EMMEANS line? If one tests
an interaction term as perhaps, for the example below,

/EMMEANS = TABLES(pcrop*nitrogen) COMPARE (nitrogen) ADJ(SIDAK)
/EMMEANS = TABLES(pcrop*nitrogen) COMPARE (pcrop) ADJ(SIDAK)

how can one know the number of tests over which the adjustment for multiple
comparisons is made? When would one choose this type of post hoc test vs. a
post hoc test such as Tukey's or Student Neuman Keuls or LSD? I know they
all control differently for the test wide error rate, but is there any other
rationale for selecting one type over another? Also, at the bottom of the
multiple comparison tests, a note always says "calculated using actual data"
(or something like that) while the EMMEANS tests say "calculated using
predicted data". How should I understand/explain the differences in
calculation of the two types of test?

I believe these questions also relate to my confusion about the plots of
"expected marginal means" that are produced with the anova commands. When
is it preferable to use those than interaction graphs produced from the
actual data?

Thank you very much.

Lucinda Tear

----- Original Message -----
From: "Marta García-Granero" <[hidden email]>
To: <[hidden email]>
Sent: Thursday, October 05, 2006 1:13 AM
Subject: Re: split plot

> Hola Julia:
>
> Saludos desde Pamplona...
>
> I have a worked example of a split-plot design but, unfortunately, a
> bit simpler than your layouts (it has only two factors, one in main
> plot an the other in sub-plots). You can use it as a starting point to
> develop your syntax. With UNIANOVA, they key is using the /TEST
> subcommand. It always works, while the approach using MIXED did not in
> some particular examples I tried.
>
> If you need more information or help, please send a small sample of
> your data and I will try to adjust the syntax for your particular
> situation.
>
> *************************************************************
> * SPLIT-PLOT ANALYSIS USING UNIANOVA *
> * AND MIXED (using EMEANS+SIDAK formultiple comparisons) *
> * Split-plot example taken from Irristat 4.4 manual: *
> * http://www.irri.org/science/software/irristat.asp *
> *************************************************************
> * RESPONSE OF MAIZE TO NITROGEN FERTILIZER *
> * WHEN IT IS GROWN AFTER DIFFERENT CROPS *
> * TALLEYRAND, H. NCRE CEREALS AGRONOMIST, IRA/GAROUA *
> * THE TRIAL WAS A SPLIT PLOT DESIGN WITH PRECEEDING CROP AS *
> * MAIN PLOT FACTOR AND FOUR LEVELS OF NITROGEN AS SUB-PLOT *
> * FACTOR: 0, 45, 90, 135 KG N/ha. MAIZE VARIETY CMS8501. *
> *************************************************************
>
> * Dataset *.
> DATA LIST FREE/block pcrop nitrogen (3 F4.0) yield (F4.2).
> BEGIN DATA
> 1 5 1 2.35 1 5 2 3.29 1 5 3 4.75 1 5 4 6.39 1 1 1 3.76 1 1 2 5.17
> 1 1 3 6.30 1 1 4 6.82 1 3 1 2.26 1 3 2 5.17 1 3 3 5.88 1 3 4 6.82
> 1 2 1 3.71 1 2 2 4.70 1 2 3 5.88 1 2 4 6.20 1 4 1 1.41 1 4 2 3.48
> 1 4 3 4.61 1 4 4 4.70 2 5 1 4.20 2 5 2 5.23 2 5 3 6.64 2 5 4 6.73
> 2 1 1 5.28 2 1 2 5.84 2 1 3 6.36 2 1 4 7.39 2 3 1 3.49 2 3 2 5.84
> 2 3 3 6.88 2 3 4 6.78 2 2 1 4.85 2 2 2 5.98 2 2 3 6.73 2 2 4 7.49
> 2 4 1 3.12 2 4 2 3.59 2 4 3 4.67 2 4 4 6.50 3 5 1 5.34 3 5 2 6.79
> 3 5 3 7.31 3 5 4 8.20 3 1 1 6.75 3 1 2 8.11 3 1 3 8.30 3 1 4 8.91
> 3 3 1 4.26 3 3 2 7.17 3 3 3 7.59 3 3 4 8.34 3 2 1 5.57 3 2 2 5.81
> 3 2 3 6.23 3 2 4 7.69 3 4 1 4.49 3 4 2 5.67 3 4 3 6.93 3 4 4 7.17
> 4 5 1 3.38 4 5 2 3.85 4 5 3 5.17 4 5 4 5.36 4 1 1 4.18 4 1 2 5.88
> 4 1 3 6.30 4 1 4 7.14 4 3 1 2.82 4 3 2 4.61 4 3 3 5.73 4 3 4 5.88
> 4 2 1 4.47 4 2 2 5.17 4 2 3 5.73 4 2 4 6.34 4 4 1 2.73 4 4 2 3.06
> 4 4 3 4.70 4 4 4 5.36
> END DATA.
>
> VALUE LABEL pcrop 1'PPEAS' 2'CROTL' 3'GNUTS' 4'CWPEA' 5'MAIZE'.
> VALUE LABEL nitrogen 1'0' 2'45' 3'90' 4'135'.
>
> * Using UNIANOVA *
>
> UNIANOVA
> yield BY block pcrop nitrogen
> /RANDOM = block
> /METHOD = SSTYPE(3)
> /INTERCEPT = EXCLUDE
> /TEST= pcrop vs block*pcrop
> /DESIGN = pcrop block block*pcrop nitrogen pcrop*nitrogen .
>
> * With MIXED, with a post-hoc analysis for main effects *.
>
> MIXED
> yield BY pcrop block nitrogen
> /FIXED = pcrop nitrogen nitrogen*pcrop | NOINT
> /METHOD = REML
> /RANDOM = block block*pcrop | COVTYPE(VC)
> /SAVE = RESID
> /EMMEANS = TABLES(pcrop) COMPARE ADJ(SIDAK)
> /EMMEANS = TABLES(nitrogen) COMPARE ADJ(SIDAK)
>
>
> j> I need help with a pair of split plot analysis.
>
> j> In the first case, I have a three factors complete randomized
> split-plot
> j> design, with one factor on whole-plot and two factors on sub-plots.
>
> j> In the second case, is also a three factors complete randomized split-
> j> plot, with two factors on whole-plot and one factor on sub-plot.
>
> j> I would like to know if SPSS can perform this kind of analysis via
> Syntax
> j> commands and how it could be done. I would be really very grateful of
> any
> j> kind of help or comment.
>
> j> Julia.
> j> PhD Student at University of Santiago de Compostela. Spain.
>
> --
> Regards,
> Dr. Marta García-Granero,PhD mailto:[hidden email]
> Statistician
> University of Navarra
> SPAIN
>
> ---
> "It is unwise to use a statistical procedure whose use one does
> not understand. SPSS syntax guide cannot supply this knowledge, and it
> is certainly no substitute for the basic understanding of statistics
> and statistical thinking that is essential for the wise choice of
> methods and the correct interpretation of their results".
>
> (Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)
>